We provide AI data classification and governance in sovereign control for sensitive communications in the regulated industries.
Internet
- Tweets 55,281
- Following 567
- Followers 4,852
- Likes 262,028
PRODUCT RELEASE: Réunion 3.7.11 on iOS.
- Various updates for the next server release
- Lay out three Pulse attachments like the web client
- Show Vivid call buttons only to users granted Vivid
- Clear Pulse notifications read on another device
- Keep Cancel and Done visible while searching room members
- Show the live server version on the Server screen
- Pulse: return the composer to the bottom once the keyboard is hidden
- Pulse: show a video poster before the file is downloaded
- Pulse: pin chats and rooms, shared with the web client
- DataRoom
- Module tab items
33
PRODUCT RELEASE: Reunion Web 3.1.112
Pulse -
New: Pulse now caches your rooms and messages so they load faster. It is on by default for testing. You can turn it off in Settings > Experimental > Pulse Cache.
New: Videos now get a preview thumbnail. When you upload a video, a thumbnail is created and shown as its preview. This works only for new videos uploaded from the web for now, and is coming to iOS and Android.
New: Pinned rooms are now saved to the server.
Improved: A new room is selected automatically after you create it.
Improved: Group creation is easier to use.
Improved: Attachments are shown in a new way. You can now right-click an image to download it.
Fixed: Large attachments now upload correctly.
Fixed: Downloaded attachments keep their original filenames instead of being renamed.
EasyCrypt -
New: Press Cmd/Ctrl + Enter to send a message.
Fixed: Pressing Tab in the subject field now moves to the message body.
Fixed: Scrolling now works on long messages.
Fixed: The message reply icon is now correct.
Data Rooms (Preview)
New: An early preview of the Data Rooms module. Turn it on in Settings > Experimental > Data Rooms. General
1
56
Most compliance failures are classification failures in disguise.
Access control, DLP, encryption, and audit logs only work if the system knows the sensitivity of the data it’s looking at.
When a regulated file leaves as a normal attachment, you lose the audit trail, the jurisdiction, and the ability to prove control. Regulators increasingly treat incomplete classification as systemic non-compliance, not a documentation gap.
Labeling is the quiet prerequisite. Get it wrong and the expensive controls become theater.
3
53
「ひと目でわかる Microsoft Purview」のDLPパートを執筆された森下さんが10/24にご登壇されます！DLP活用を検討されている方はぜひご参加ください🚀
📣今月は24日(土)お昼の開催です！
「ひと目でわかるMicrosoft Purview」のDLP関連の章を書かれた森下さんに、DLP設定後の気になるポイント🔍をお話いただきます！
ゆるっとPurview勉強会 #5
purview.connpass.com/event/4… #ゆるPurview
1
1
7
539
Copilot Cowork handles new requests mid-run. Add meeting scheduling or an email update partway through and it integrates them into the active plan. Check it out. youtu.be/4vkPv9lX64k
Access Anthropic and OpenAI models directly from Microsoft 365 Copilot. Generate briefing documents, presentations, and Excel files from a single prompt with Copilot Cowork, pulling from your emails, calendar, and SharePoint through Work IQ — and fold in new tasks mid-run without stopping.
Using Copilot Cowork, you can use the same platform that powers Claude Cowork. It’s designed for long-running, multi-step task automation. #MultiModelCopilot #Microsoft365Copilot #AIModels #GenerativeAI #ArtificialIntelligence
2
1
7
2,221
Let’s say you’re working in Visual Studio Code on a healthcare app, and you select a few lines of code to debug a query—a routine moment in your day. That snippet might include connection strings, test data with real patient info, and part of your schema. You ask Copilot to help and approve an MCP tool that connects to a remote server—and all of it gets sent to external servers. That’s not just risky. It could be a compliance violation under HIPAA, SOX, or PCI-DSS, depending on what gets transmitted.
From my @OReillyMedia Radar article, “MCP Introduces Deep Integration—and Serious Security Concerns”
2
1
76
ICYMI: Your employees are already pasting sensitive data into AI tools you never approved. Here's how to inventory shadow AI apps, see what's at risk with Microsoft Purview, and enforce DLP: adamtheautomator.com/stop-sh…
2
8
72
6,782
🦔Meta's new AI agent, Muse, asks people to connect their email and bank accounts so it can shop, book travel and pay bills for them. 404 Media reports Meta engineers rushed to patch a flaw two weeks before launch that could have let an ordinary Muse user reach into Meta's internal databases, and some senior engineers still expect a major breach.
My Take
I wouldn't link a bank account to any AI agent right now, Meta's or anybody else's. Muse launched in September and this is at least its third security story. Two developers got it to hand over its own file system, and The Information reported a separate bug that could have exposed a user's emails and files. Meta says the file export wasn't a breach and that it fixed the bug.
Meta expects to spend as much as $145 billion on AI this year, and investors want to see it pay off. An agent with your card on file makes money every time it books a flight, so I understand the hurry. But every AI model makes mistakes you can't predict, and anyone who slips the right instructions into an email or webpage can talk it into things. Put that together with access to your checking account and one bad day gets expensive fast. Muse got about 2.8 million downloads in its first two weeks. If you already connected your bank, I'd reconsider that decision...
Hedgie🤗
52
118
413
21,848
改正個人情報保護法はEUの保護水準を満たしているのか？
医師・弁護士らがEUに意見書提出し十分性認定の是非を問いました。
10月6日記者会見
youtu.be/VAv17f8UsK4?si=H0_Y…
1
91
182
5,463
今年4月、国会で「Mythos級の攻撃能力を持つモデルが拡散するまで早ければ数ヶ月しかない」と申し上げておりました。
実際、8月末に公開されたGLM-5.3の能力はほぼMythos級のようで、実際に発言が実現してしまった形になります。
データの漏洩事故も相次いでおり、対応を急ぐ必要があります
本日の総務委員会で、Claude Mythos等サイバー攻撃能力の高いモデルの登場に伴い、喫緊の対応が必要ではないかと課題提起をいたしました。
チームみらいはMythosの件について、発表直後から繰り返し、様々な委員会にて課題提起をしております。当初4/10段階ではまだ「現在議論を深めている」（外務委・参考人）という回答だったものの、徐々に温度感も上がり、本日は総務大臣より「喫緊の課題だと認識」との答弁がありました。
AIの開発競争は熾烈で、Anthropic社以外にもこのレベルのモデルを開発する会社は今後多く現れると思います。攻撃能力が拡散するまで下手すると数ヶ月〜1年程度しか猶予がないかもしれません。こちら非常に重要かつスピードが求められる論点ですので、今後も引き続き取り組みます
（参考）チームみらいのMythosの課題提起に対する政府答弁推移
4/10 政府参考人（国家サイバー統括室）「議論を深めている」
4/15 官房長官「最新の技術動向にも注視」
4/15 AI担当大臣「危機感は非常に共有」
4/17 総理「適時に必要な対応を」
4/21 総務大臣「喫緊の課題だと認識」← New
392
1,587
9,154
2,492,510
The Wikimedia Foundation says rogue OpenAI agents made unauthorized Wikipedia edits and may have partly caused a May outage.
bleepingcomputer.com/news/se…
13
12
59
8,542
Sam Altman is happy to accept some bad things happening. As long as they happen to you.
POLITICO
13
40
290
11,719
Exercise and remission from cancer: A randomized trial found 37% improved survival, 28% disease-free survival. But little is being done to incorporate in practice except @UPMC 1st health system covering "exercise oncology" program.
gift link, by @AgrawalNina
nytimes.com/2026/10/06/well/…
28
226
765
69,444
AI and Challenges in Workforce Development
Train the humans and machines together.
philvenables.com/post/ai-and…
2
1
4
577
Open-source implementation of the entire Adobe Suite, reverse engineered by AI.
Closed source is dead.
This is the doom I predicted a few days ago, coming for Photoshop. A clean-room open-source reimplementation.
No prizes for guessing that they decompiled Photoshop to source code, processed that to some kind of non-code specification language, then fed the spec to an LLM with an instruction to generate Rust.
Adobe just got nuked. And closed source is dead, dead, dead.
github.com/storytold/photocr…
22
61
762
51,306
Let me spell out what this means for privacy:
Surveillance is hidden all through opaque software. You are never meant to find it.
But AI can decompile, analyze, & expose exactly what your apps are doing.
The age of "trust us" software is ending. Privacy revolution incoming.
Open-source implementation of the entire Adobe Suite, reverse engineered by AI.
Closed source is dead.
58
297
1,984
33,473